引導式測驗 · 中等 等級
運動檢測測驗的背景扣除
測試固定相機背景模型、前景遮罩、更新率權衡、陰影、相機運動和任務級錯誤。
問題 1 的 8
What does a foreground mask from background subtraction primarily indicate?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- What does a foreground mask from background subtraction primarily indicate?
- The identity of each moving person
- The object class of every connected region
- Pixels that differ from the estimated background
- A metrically accurate 3D map
- Why can an overly fast background update miss a person who stops?
- A stopped person always leaves the camera view
- The background never changes after initialization
- Shadow logic must remove stationary people
- The stationary person may be absorbed into the background model
- What can an overly slow update do after a vehicle parks?
- Immediately learn the parked vehicle as background
- Continue flagging the new stationary vehicle as foreground
- Guarantee perfect separation of shadows
- Prevent all camera motion artifacts
- Why can a moving camera break basic background subtraction?
- Most pixels shift even when world objects remain still
- A moving camera cannot produce video
- The model automatically knows every object’s identity
- Motion converts each pixel into depth
- Wind-blown leaves trigger many mask pixels. Which conclusion is justified?
- A target animal must be present
- Every pixel change is a labeling error
- Scene motion was detected, not necessarily a target animal
- The mask reveals intent of moving objects
- How can shadows affect foreground masks?
- They always improve object counts
- They alter image intensity and can enlarge or create apparent motion regions
- They directly identify a class label
- They prevent background updates
- Two people overlap and become one blob. What limitation does this show?
- A foreground region is not automatically one tracked object
- A connected mask always represents one person
- Overlapping people cannot be observed
- MOG2 assigns persistent identities
- Which test best assesses a motion alarm built on subtraction?
- Only a visually clean mask screenshot
- Only processing time per frame
- Only total background-pixel count
- False alarms and missed targets across real lighting and activity